asciilifeform: problem comes if you want to run non-contemporary proggies on it ( musl, gnat 4.9.x, etc . ) linus permitted abi to change.☟︎
asciilifeform: the beauty of synthetic 'iron' is that you can actually use just about any kernel, because there is no driver gnarl, you're already baking the drivers with own hands
asciilifeform: it was ~2.4~ that some folx are keeping alive, cuz it would fit, with trimming, into <1M. but if yer using 2.4.x you will need a contemporary userland.
asciilifeform: never found any reason to keep alive 2.6.x, hand-cut later kernel can typically sit down in same footprint even in small irons
asciilifeform: mp_en_viaje: i don't have anything running a 2.x, and haven't for some years ( ye olde 'wrt54g' was i think the last piece of iron i had which ran one )
asciilifeform: mod6: re backups -- if one absolutely must make a hand-cranked copy of a node, the correct method is 'dumpblock' (on donor end) and 'eatblock' (on recipient) , dun require taking down either.
asciilifeform: for the smaller boxen (various arm, 'pcengine', similar) you can actually put a 12v lead-acid in line with the dc connector. and never suffer 'oops, mains flicker' db corrupt.
asciilifeform: http://btcbase.org/log/2019-07-16#1922900 << i dunthink i've ever witnessed the corrupted db (tho i believe the folx who have.) possibly because i have never run node off anything other than always-on sinusoidal inverter ups, tho☝︎☟︎
asciilifeform: realtime auto-backup, what better backup could anything ever have.
asciilifeform: http://btcbase.org/log/2019-07-16#1922895 << recall that mp_en_viaje prescribed that the wallet oughta be sawed off into own proggy. imho this is entirely Right Thing. but no one has yet found the free hands with which to do, afaik☝︎☟︎
asciilifeform: aaand gcc broke support around same time.
asciilifeform: ... and mips in general. ( how do you suppose the ball-of-shit arm arch became dominant. to this day chinese pay tribute to britain for 'permission' to make arm. why ? at one time mips had patent, but expired in 2009 or so. and in march of '19 orig. mips verilog entirely opensoresd and made public... yet 0 fast mips on the market. guess why. )☟︎
asciilifeform: near as i can guess, this was done in order to help kill loongson sinomips.☟︎
asciilifeform: this wouldn't even be interesting ('hey use pre-castration snapshot') except that this snip pre-dates addition of musl to buildroot.
asciilifeform: meanwhile, in shitgnomery, '2013.11-rc3, Released November 26th, 2013 Fixes all over the tree. Architecture: Mark MIPS I, II, III and IV as deprecated.' ('buildroot' www)
asciilifeform: mod6: as i parsed it, she already is using tunnel.
asciilifeform: fwiw it lies down cleanly on errything 3.x--present, afaik
asciilifeform: re that kernel -- someone gotta genesis a kernel. ( if no one has any constructive input re ~which~ one, then i will, and it will be somewhat arbitrary. ) atm that patch is a bad-old-style patch, rather than vpatch.☟︎
asciilifeform: in re ffa : ch20 and most of 21 written. but all of it paused atm , making detailed map/plan.
asciilifeform: the arch illustrated by the sim is simple enuff that one could fit it in, e.g., even very modest fpga. which means that nao we have reproducible kernel, gcc, toolchain, etc. for machine that can be baked , if needed, on demand, in whatever qty.
asciilifeform: gcc 4.9.x apparently WILL emit mips-r2-istic instrs no matter what flags are given.☟︎
asciilifeform: ditto in buildroot, some time (remains to be established) b/w 2013 ver and current, also nixed mips1
asciilifeform: also in the process made sad discovery : previously asciilifeform always pronounced 'gcc 4.9 is usable' but! turns out some time b/w 4.8.1 and 4.9 , someone silently sabotaged support for classical mips1☟︎
asciilifeform: so ended up learning from 0 , how to port linux kernel to a never-seen 'machine'
asciilifeform: orig. author of 'cmips' btw was of 0 use, i wrote to him a while back and not long ago he answerd 'i lost hdd, lost errything'☟︎
asciilifeform: re : honest 'virtual hosts' , and re asciilifeform's backlog of outputs to be published : i have a working kernel and userland for that mips emu . runs as buildroot, a la 'rotor' . will be featured in conveyor article.
asciilifeform: possibly i was thinking about 'bitvps'
asciilifeform: hrm i thought in 'linode' , operator succumbed to alcoholism
asciilifeform: i admit i'm curious now, what kinda budget is it where <100 u.s/mo is 'high' . and where exactly girlattorney has been getting physical box colo that costs substantially less.☟︎
asciilifeform: there's a vacant rk, prolly adequate for anything short of trb node
asciilifeform: speaking of which, from old compendium of 'militarisms' : new division commander shows up, and , to the troops lined up on platz, utters : 'i will discipline just like you know from kindergarten : some to tribunal, some will be shot in front of the regiment... '
asciilifeform: asciilifeform, for example, in usa. where for 5k you can not even usually have the water pipe to your house fixed.
asciilifeform: girlattorney: you did not specify your geography, so i assumed usa (not because always assume usa, but because it is country with world's worst bandwidth market. as in, behind nigeria, congo, et al )
asciilifeform knows several people stuck on what may as well be gas modem
asciilifeform: diana_coman: lolyes. i also like 'gas modem'.
asciilifeform: tunnel method worx, but imposes considerably overhead, girlattorney , you will want to get a serious pipe. i don't imagine it is a painful expense for attourney
asciilifeform: instead get working linux and toolchain on the arm box, then build w/ mod6's system.
asciilifeform: no need to cross compile unless you're building for something too small to run gcc itself ( and chances are it won't run trb , if so small , as with 'pogo' )
asciilifeform: i.e. repeatable build, doesn't care what's on your box currently
asciilifeform: girlattorney: mod6's build system ( based on asciilifeform's earlier 2015 'rotor' (see logs) ) builds first a frozen-in-amber gcc & friends, ~then~ the depds (db etc) , ~then~ trb per se
asciilifeform: if you have a working toolchain now, you can build and be up & running in 40min. or so.
